Video summary

Agent sessions and where agents run | Ep 4 of 6 - VS Code Learn

Main summary

Key takeaways

Technology

Key concepts and product features

Session management in VS Code

  • View, switch between, fork, archive, rename, and delete agent sessions from the session view.
  • Create new sessions using the + button or the “new session” UI.
  • Right-click a session for quick organization or cleanup actions.

Where agents can run

  • Local: agent work runs on your machine (used in the demo).
  • Copilot CLI: runs a background agent locally via the GitHub Copilot CLI (terminal-first workflow).
  • Cloud (GitHub platform): compute runs on github.com, enabling asynchronous runs independent of your local machine.
  • Third-party agents: option to integrate other agent providers.

Session modes

  • Plan mode: agent acts as a planner (outlines steps, supports iterative planning).
  • Ask mode: chatbot-style conversational mode (good for initial exploration).
  • Agent mode: executes work (makes changes, opens PRs, etc.).

Recommendation: iterate using Ask/Plan modes, then switch to Agent mode when ready to implement.

Example workflow demonstrated

  1. Use “changes” as context so the agent sees modified files.
  2. Commit and push changes locally.
  3. Start a cloud session to “write a README”; Copilot creates a work-in-progress pull request on GitHub.
  4. Inspect the running session and the PR in VS Code and on github.com (Agents page).
  5. From GitHub you can view session progress, the model used, premium request counts, and tool costs.

Visibility and controls

  • Sessions started in VS Code appear in the sessions list; hovering reveals where they run (local vs cloud).
  • The GitHub Agents UI mirrors VS Code session info and provides cost/usage details.
  • Multiple sessions can run concurrently; you can switch focus between them.

Tutorials / guides mentioned (to be covered in other episodes)

  • Deep dive video(s) on GitHub Copilot CLI.
  • Customizing agents / custom instructions.
  • Next episode: inspecting agent calls/actions, understanding LLM calls and tool usage.

Practical takeaways

  • You can run and monitor agents locally, via Copilot CLI, or in the GitHub cloud.
  • Use Plan/Ask modes to design tasks and Agent mode to execute them.
  • The platform provides tracing and visibility (model, requests, tool costs) for sessions and PRs.
  • This setup is useful for running asynchronous work on the platform while continuing local development.

Main speaker / sources

  • VS Code Learn series instructor (narrator of the tutorial)
  • Platforms referenced: Visual Studio Code, GitHub Copilot, GitHub (Agents UI), GitHub Copilot CLI, and third-party agent integrations.

Rate this summary

Your feedback will help improve summaries.

Improve this summary

Reprocess with a stronger model when the summary feels incomplete or inaccurate.

Pro

Translate summary in another language

Pro

Ask questions to this video

Chat for follow-up questions, clarifications, and source-backed answers.

Coming soon

Share this summary

Original video